AlphaFold 2 and GitHub Copilot

AI solves a 50-year scientific problem

In July 2021, DeepMind published AlphaFold 2 in Nature — a system that could predict the three-dimensional structure of a protein from its amino acid sequence with near-experimental accuracy. The protein folding problem had been one of the grand challenges of biology for 50 years: knowing how a protein folds determines its function, and understanding this is crucial for drug development, disease research, and synthetic biology. AlphaFold 2 essentially solved it.

The scale of the impact

DeepMind subsequently made the AlphaFold database publicly available, initially with structures for 350,000 proteins and later expanding to nearly all known proteins — more than 200 million structures. This accelerated research across biology and medicine on a scale unprecedented in the history of AI. The Nobel Prize in Chemistry 2024 was awarded to Demis Hassabis and John Jumper (DeepMind) and David Baker for their work on protein structure prediction.

GitHub Copilot: AI for developers

In June 2021, GitHub and OpenAI launched GitHub Copilot — an AI coding assistant built on Codex, a version of GPT-3 fine-tuned on public code. Copilot could autocomplete entire functions, generate boilerplate code, and suggest implementations based on comments in natural language. For software developers, it was the first genuinely useful AI tool in their daily workflow. Within a year it had millions of users and became the template for a new category of developer tools.

Sources

• Jumper, J. et al. (2021). Highly accurate protein structure prediction with AlphaFold. Nature, 596, 583–589.
• Wikipedia — AlphaFold
• Wikipedia — GitHub Copilot